In this paper, we propose a simple way to pass borders of traditional text data augmentation methods. Machine learning and deep learning are compared using SVM and CNN-BiLSTM model pipeline. Moreover, it includes the comparison of statistical and contextual methods, Word2Vec and BERTurk respectively. Also, normalization effects are investigated using Zemberek codes.
Codes folder provides the task related codes.
Dataset folder includes the description of the proposed dataset.
Figures folder contains representation of dataset gathering process and illustration of BERTurk-CNN-BiLSTM model pipeline.
T. Tanyel, B. Alkurdi and S. Ayvaz, "Linguistic-based Data Augmentation Approach for Offensive Language Detection," 2022 7th International Conference on Computer Science and Engineering (UBMK), 2022, pp. 1-6, doi: 10.1109/UBMK55850.2022.9919562.